[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fYSN130L12qlanB-WO4FmR0L6CRkLDVB4cENTW1Ody5w":3},[4,16,30,42,52,60,75,89,101,111,119,127,138,148,159,165,175,185,193,197,203,209,216,220,224,231,237,246,252,257],{"name":5,"provider":6,"releaseDate":7,"contextWindow":8,"contextWindowNum":9,"pricing":10,"arena":12,"coding":7,"reasoning":7,"speed":7,"openSource":13,"highlights":14,"lastVerified":15},"Claude Opus 4.7","Anthropic",null,"128K",0,{"display":11,"input":9,"output":9},"",1567,false,"Opus 4.7 features a new tokenizer that inflates token counts by 35-45%.","2026-06-08",{"name":17,"provider":18,"releaseDate":19,"contextWindow":20,"contextWindowNum":21,"pricing":22,"arena":26,"coding":27,"reasoning":27,"speed":28,"openSource":13,"highlights":29,"lastVerified":15},"GPT-5.4","OpenAI","2026-03","1M",1000000,{"display":23,"input":24,"output":25},"$3\u002F$15",3,15,1560,"unknown","~50 t\u002Fs","Unifies Codex + GPT; 1M context; built-in computer use",{"name":31,"provider":6,"releaseDate":32,"contextWindow":20,"contextWindowNum":21,"pricing":33,"arena":37,"coding":38,"reasoning":39,"speed":40,"openSource":13,"highlights":41,"lastVerified":15},"Claude Opus 4.6","2026-02",{"display":34,"input":35,"output":36},"$5\u002F$25",5,25,1549,"80.8% SWE-bench","65.4% Terminal-Bench","~40 t\u002Fs","#1 Arena Hard Prompts & Coding; 128K max output",{"name":43,"provider":44,"releaseDate":19,"contextWindow":20,"contextWindowNum":21,"pricing":45,"arena":49,"coding":27,"reasoning":27,"speed":50,"openSource":13,"highlights":51,"lastVerified":15},"Gemini 3.1 Flash Lite","Google",{"display":46,"input":47,"output":48},"$0.1\u002F$0.4",0.1,0.4,1492,"~200 t\u002Fs","#3 Arena overall; #1 creative writing; ultra-fast",{"name":53,"provider":6,"releaseDate":32,"contextWindow":20,"contextWindowNum":21,"pricing":54,"arena":55,"coding":56,"reasoning":57,"speed":58,"openSource":13,"highlights":59,"lastVerified":15},"Claude Sonnet 4.6",{"display":23,"input":24,"output":25},1440,"79.6% SWE-bench","72.5% OSWorld","~80 t\u002Fs","Best value frontier; beats Opus 4.5 in 59% head-to-head",{"name":61,"provider":62,"releaseDate":63,"contextWindow":8,"contextWindowNum":64,"pricing":65,"arena":69,"coding":70,"reasoning":71,"speed":72,"openSource":73,"highlights":74,"lastVerified":15},"Qwen 3 235B","Alibaba","2025-04",128000,{"display":66,"input":67,"output":68},"$0.86\u002F$2",0.86,2,1422,"70.7% LiveCodeBench","2056 CodeForces ELO","~65 t\u002Fs",true,"235B MoE (22B active); Apache 2.0; strongest OSS competitive programming",{"name":76,"provider":77,"releaseDate":78,"contextWindow":79,"contextWindowNum":80,"pricing":81,"arena":85,"coding":27,"reasoning":86,"speed":87,"openSource":73,"highlights":88,"lastVerified":15},"Mistral Large 3","Mistral","2025-12","256K",256000,{"display":82,"input":83,"output":84},"$0.5\u002F$1.5",0.5,1.5,1418,"43.9% GPQA Diamond","~70 t\u002Fs","675B MoE (41B active); Apache 2.0; best cost-efficiency frontier",{"name":90,"provider":44,"releaseDate":91,"contextWindow":20,"contextWindowNum":21,"pricing":92,"arena":96,"coding":97,"reasoning":98,"speed":99,"openSource":13,"highlights":100,"lastVerified":15},"Gemini 2.5 Pro","2025-03",{"display":93,"input":94,"output":95},"$1.25\u002F$10",1.25,10,1412,"75.6% LiveCodeBench","84.6% GPQA Diamond","~60 t\u002Fs","Thinking model; top WebDev Arena 1415; native multimodal",{"name":102,"provider":18,"releaseDate":63,"contextWindow":103,"contextWindowNum":104,"pricing":105,"arena":108,"coding":27,"reasoning":27,"speed":109,"openSource":13,"highlights":110,"lastVerified":15},"o3","200K",200000,{"display":106,"input":95,"output":107},"$10\u002F$40",40,1402,"~30 t\u002Fs","Strongest OpenAI reasoning model",{"name":112,"provider":6,"releaseDate":113,"contextWindow":103,"contextWindowNum":104,"pricing":114,"arena":115,"coding":116,"reasoning":27,"speed":117,"openSource":13,"highlights":118,"lastVerified":15},"Claude Opus 4.5","2025-11",{"display":34,"input":35,"output":36},1380,"80.9% SWE-bench","~35 t\u002Fs","Major price cut from Opus 4; strong agentic coding",{"name":120,"provider":121,"releaseDate":32,"contextWindow":8,"contextWindowNum":64,"pricing":122,"arena":115,"coding":27,"reasoning":27,"speed":58,"openSource":73,"highlights":126,"lastVerified":15},"DeepSeek V3.2","DeepSeek",{"display":123,"input":124,"output":125},"$0.14\u002F$0.42",0.14,0.42,"~90% GPT-5.4 quality at 1\u002F50th cost; best value model",{"name":128,"provider":129,"releaseDate":130,"contextWindow":8,"contextWindowNum":64,"pricing":131,"arena":115,"coding":135,"reasoning":136,"speed":28,"openSource":73,"highlights":137,"lastVerified":15},"Kimi K2","Moonshot","2025-07",{"display":132,"input":133,"output":134},"$0.55\u002F$2.2",0.55,2.2,"65.8% SWE-bench","60.2% BrowseComp","1T params; Agent Swarm (100 agents); Modified MIT",{"name":139,"provider":121,"releaseDate":140,"contextWindow":8,"contextWindowNum":64,"pricing":141,"arena":144,"coding":27,"reasoning":145,"speed":146,"openSource":73,"highlights":147,"lastVerified":15},"DeepSeek R1","2025-01",{"display":142,"input":133,"output":143},"$0.55\u002F$2.19",2.19,1363,"#1 Math & Coding Arena","~45 t\u002Fs","671B MoE (37B active); MIT license; distilled variants available",{"name":149,"provider":150,"releaseDate":151,"contextWindow":152,"contextWindowNum":153,"pricing":154,"arena":155,"coding":27,"reasoning":156,"speed":157,"openSource":13,"highlights":158,"lastVerified":15},"Grok 3","xAI","2025-02","131K",131000,{"display":23,"input":24,"output":25},1350,"93.3% AIME 2025","~55 t\u002Fs","Strong math\u002Fscience; now legacy (Grok 4 series launched)",{"name":160,"provider":150,"releaseDate":161,"contextWindow":79,"contextWindowNum":80,"pricing":162,"arena":163,"coding":27,"reasoning":27,"speed":146,"openSource":13,"highlights":164,"lastVerified":15},"Grok 4","2026-01",{"display":34,"input":35,"output":36},1340,"Top-5 Arena; strong reasoning & real-time X data",{"name":166,"provider":6,"releaseDate":167,"contextWindow":103,"contextWindowNum":104,"pricing":168,"arena":172,"coding":27,"reasoning":27,"speed":173,"openSource":13,"highlights":174,"lastVerified":15},"Claude Haiku 4.5","2025-10",{"display":169,"input":170,"output":171},"$0.8\u002F$4",0.8,4,1290,"~120 t\u002Fs","Fastest Claude, cheapest tier",{"name":176,"provider":18,"releaseDate":177,"contextWindow":8,"contextWindowNum":64,"pricing":178,"arena":181,"coding":182,"reasoning":27,"speed":183,"openSource":13,"highlights":184,"lastVerified":15},"GPT-4o","2024-05",{"display":179,"input":180,"output":95},"$2.5\u002F$10",2.5,1285,"30.8% SWE-bench","~100 t\u002Fs","Legacy but still available; superseded by GPT-5 family",{"name":186,"provider":44,"releaseDate":91,"contextWindow":20,"contextWindowNum":21,"pricing":187,"arena":190,"coding":27,"reasoning":27,"speed":191,"openSource":13,"highlights":192,"lastVerified":15},"Gemini 2.5 Flash",{"display":188,"input":189,"output":180},"$0.30\u002F$2.50",0.3,1268,"~150 t\u002Fs","Cheapest frontier model at scale",{"name":194,"provider":44,"releaseDate":7,"contextWindow":8,"contextWindowNum":9,"pricing":195,"arena":7,"coding":7,"reasoning":7,"speed":7,"openSource":13,"highlights":196,"lastVerified":15},"Gemini 3.2 Flash",{"display":11,"input":9,"output":9},"Faster than Gemini 3.1 Pro with improved performance.",{"name":198,"provider":199,"releaseDate":32,"contextWindow":79,"contextWindowNum":80,"pricing":200,"arena":7,"coding":27,"reasoning":27,"speed":87,"openSource":73,"highlights":202,"lastVerified":15},"MiMo V2","Xiaomi",{"display":201,"input":9,"output":9},"Free","Free coding model; 256K context; open weights",{"name":204,"provider":205,"releaseDate":7,"contextWindow":206,"contextWindowNum":9,"pricing":207,"arena":7,"coding":7,"reasoning":7,"speed":7,"openSource":13,"highlights":208,"lastVerified":15},"Muse Spark","Meta","N\u002FA",{"display":11,"input":9,"output":9},"Muse Spark achieves its reasoning capabilities using over an order of magnitude less compute than Llama 4 Maverick.",{"name":210,"provider":205,"releaseDate":63,"contextWindow":20,"contextWindowNum":21,"pricing":211,"arena":7,"coding":27,"reasoning":27,"speed":99,"openSource":73,"highlights":215,"lastVerified":15},"Llama 4 Maverick",{"display":212,"input":213,"output":214},"$0.15\u002F$0.60",0.15,0.6,"400B MoE (17B active); strong multimodal; open weights",{"name":217,"provider":18,"releaseDate":7,"contextWindow":8,"contextWindowNum":9,"pricing":218,"arena":7,"coding":7,"reasoning":7,"speed":7,"openSource":13,"highlights":219,"lastVerified":15},"GPT-5.5",{"display":11,"input":9,"output":9},"OpenAI's latest model, GPT-5.5, offers advanced capabilities for coding and complex tasks.",{"name":221,"provider":121,"releaseDate":7,"contextWindow":8,"contextWindowNum":9,"pricing":222,"arena":7,"coding":7,"reasoning":7,"speed":7,"openSource":13,"highlights":223,"lastVerified":15},"DeepSeek V4-Pro",{"display":11,"input":9,"output":9},"DeepSeek V4-Pro offers a significant price reduction, making it one of the most cost-effective options in the market.",{"name":225,"provider":121,"releaseDate":7,"contextWindow":8,"contextWindowNum":9,"pricing":226,"arena":7,"coding":7,"reasoning":7,"speed":7,"openSource":73,"highlights":230,"lastVerified":15},"DeepSeek V4",{"display":227,"input":228,"output":229},"$0.435\u002F$0.87",0.435,0.87,"A cost-focused DeepSeek release that recent leaderboards describe as a strong value option for reasoning and general use.",{"name":232,"provider":6,"releaseDate":7,"contextWindow":103,"contextWindowNum":9,"pricing":233,"arena":7,"coding":7,"reasoning":7,"speed":7,"openSource":13,"highlights":236,"lastVerified":15},"Claude Opus 4.8",{"display":234,"input":25,"output":235},"$15\u002F$75",75,"Improves coding, reasoning, reliability, and agentic workflows while keeping standard API pricing unchanged from Opus 4.7.",{"name":238,"provider":205,"releaseDate":63,"contextWindow":239,"contextWindowNum":240,"pricing":241,"arena":7,"coding":27,"reasoning":27,"speed":244,"openSource":73,"highlights":245,"lastVerified":15},"Llama 4 Scout","10M",10000000,{"display":242,"input":243,"output":189},"$0.08\u002F$0.3",0.08,"~90 t\u002Fs","10M context industry record; 109B MoE (17B active)",{"name":247,"provider":44,"releaseDate":7,"contextWindow":20,"contextWindowNum":9,"pricing":248,"arena":7,"coding":7,"reasoning":7,"speed":7,"openSource":13,"highlights":251,"lastVerified":15},"Gemini 3.5 Flash",{"display":249,"input":84,"output":250},"$1.50\u002F$9",9,"GA on May 19, 2026; faster than prior frontier models and strong on coding and agentic benchmarks.",{"name":253,"provider":254,"releaseDate":7,"contextWindow":20,"contextWindowNum":9,"pricing":255,"arena":7,"coding":7,"reasoning":7,"speed":7,"openSource":73,"highlights":256,"lastVerified":15},"MiniMax M3","MiniMax",{"display":11,"input":9,"output":9},"Open-weight multimodal model with a one-million-token context window and strong coding performance.",{"name":258,"provider":77,"releaseDate":161,"contextWindow":79,"contextWindowNum":80,"pricing":259,"arena":7,"coding":27,"reasoning":27,"speed":183,"openSource":73,"highlights":263,"lastVerified":15},"Devstral 2",{"display":260,"input":261,"output":262},"$0.05\u002F$0.22",0.05,0.22,"Cheapest agentic coding model; 256K context"]